Made with ML - Design, develop, deploy ML applications

In today’s newsletter:

• Mino: Build Web AI agents just using natural language

• Made with ML: Design, develop, deploy and iterate on production-grade ML applications

• DeepCode: Open Source Agentic Coding Framework

Reading time: 3 minutes.

Build Web AI Agents Using Natural language

Many workflows depend on websites that do not expose APIs. These sites often sit behind authentication, rely heavily on client-side JavaScript, and change frequently making traditional scraping fragile and expensive to maintain.

Most browser-based agents rely on continuous screenshot-based reasoning. While flexible, this approach introduces high latency, increased inference cost, and non-deterministic behavior across runs.

Mino takes a different approach. It uses model reasoning once to understand a workflow, then compiles that understanding into a fixed execution path. Subsequent runs execute directly without repeated reasoning resulting in faster, more predictable automation.

What Mino supports:

• Authenticated sessions and logged-in workflows

• Dynamic, multi-step interactions

• Parallel execution across multiple pages

• Deterministic replays after initial compilation

This makes Mino well-suited for stable, repeatable browser workflows where reliability and cost matter more than continuous reasoning.

Made with ML

Made with ML is a comprehensive guide focused on building production-grade machine learning systems, not just training models.

Instead of treating ML as an isolated modeling task, the guide covers the entire lifecycle from problem formulation to deployment and iteration through a software engineering lens..

What it covers:

• First-principles understanding of core ML concepts

• Applying software engineering best practices to ML development

• Scaling data pipelines, training, tuning, and serving in Python

• End-to-end MLOps: tracking, testing, orchestration

• Promotion workflows from development to production without infra changes

• CI/CD pipelines for continuous training and modular deployment

A solid reference for engineers moving from notebooks to maintainable ML systems.

🔗 Check out the Github Repo

Open Source Agentic Coding Framework

DeepCode is an agentic coding framework that converts research papers, text prompts, and URLs into production-ready codebases using multi-agent orchestration.

Instead of generating isolated snippets, DeepCode builds complete systems covering algorithms, backend services, and frontend applications.

Core pipelines:

• Paper2Code - Translates academic algorithms into reproducible implementations

• Text2Web - Generates complete frontend applications from natural language

• Text2Backend - Produces scalable backend services from requirements

Key features:

• Multi-modal input: papers, PDFs, DOCs, PPTs, HTML, and URLs

• Context-aware code generation using dependency graphs and CodeRAG

• Automated quality checks: static analysis, test generation, documentation

• Efficient handling of long documents via intelligent segmentation

It’s 100% open source.

🔗 Check out the Github Repo

That’s a Wrap

That’s all for today. Thank you for reading today’s edition. See you in the next issue with more AI Engineering insights.

PS: We curate this AI Engineering content for free, and your support means everything. If you find value in what you read, consider sharing it with a friend or two.

Your feedback is valuable: If there’s a topic you’re stuck on or curious about, reply to this email. We’re building this for you, and your feedback helps shape what we send.

WORK WITH US

Looking to promote your company, product, or service to 160K+ AI developers? Get in touch today by replying to this email.